A useful robot is one that fulfils its intended function. In a factory setting, where robots have been used successfully for decades, this function is often singular and clearly defined. Similarly, the surroundings of the robot are mostly known, sterile, and unobstructed. Taking robotic systems out of these conditions and into the real world comes with numerous challenges that are non-existent in factory cages. We want personal robots to cope with the uncertain and dynamic environments we inhabit, while at the same time managing and solving diverse tasks. Reconfigurable robots aim to achieve this by changing shape and function to address a variety of applications, environments, and users. While reconfigurable robots carry a lot of promise, finding a balance between the system's adaptability, the extent to which it can alter shape and function, and the added complexity is difficult. Research efforts have largely focused on proof-of-concept studies with limited reconfigurability and application range, avoiding the increasingly overwhelming mechanical, computational, and electronic complexities. This thesis introduces a new paradigm to the world of reconfigurable robotics with an inherent adaptability through simplification of the underlying structure. Approximating physical structures through polygon abstractions, similar to computer graphics, such systems can assume a wide range of structural or functional three-dimensional shapes. Based on this paradigm, it also presents a new robotic platform combining the concepts behind both modular and origami robotics, as well as reconfigurable mechanisms and polygon meshing. In order to take advantage of this new paradigm, a diverse set of problems must be investigated, spanning multiple robotic disciplines. With an increasing degree of reconfigurability, both within a module and the overall system, the growing physical and mechatronic requirements need to be analysed and addressed accordingly. New reconfiguration algorithms and control strategies need to be developed to cope with the large, and constantly changing, number of degrees of freedom. These must then be synchronised and scaled appropriately, leveraging modularity at multiple levels, to accomplish diverse sets of tasks and functions. Addressing the challenges associated with this new robotic paradigm and proving its viability provides the context for this thesis. In a first phase it outlines the initial conception, studying scalability and applicability through the combination of modularity and origami robots with a first prototype and its use in multiple scenarios. It continues with the development and analysis of several building blocks of modular origami robots, both mechanical and algorithmic, analysing mechanisms for the coupling between modules and the reconfiguration process. In the second phase the proposed paradigm is elaborated into its full form, integrating and examining reconfigurable mechanisms and polygon meshing. The resulting morphological and functional flexibility is validated through the development and testing of a highly sophisticated modular robotic system. Individual modules can alter their own triangular shape, drive towards and attach to each other, and transform into functional three-dimensional configurations. The conceptual and physical systems developed and studied in this thesis answer some of the challenges posed by this new paradigm and underline the potential of reconfigurability in robotics.
We derive an algorithm of optimal complexity which determines whether a given matrix is a Cauchy matrix, and which exactly recovers the Cauchy points defining a Cauchy matrix from the matrix entries. Moreover, we study how to approximate a given matrix by a Cauchy matrix with a particular focus on the recovery of Cauchy points from noisy data. We derive an approximation algorithm of optimal complexity for this task, and prove approximation bounds. Numerical examples illustrate our theoretical results.
Metal-Organic Frameworks (MOFs) are a class of porous materials that are applicable in many energy and environmentally relevant areas, due to their unique features including unprecedented internal surface areas and easy chemical tunability. Gas separation and storage are among the most important applications for which MOFs have been extensively studied. Considering the large number of potential MOFs that can be accessed, through a combination of numerous metal clusters and organic linkers, determining the relationship between their structural features (such as pore size and shape and chemical functionalities) and corresponding gas adsorption properties is of utmost importance. In this regard, in-situ characterization, especially diffraction, can be advantageous. MOFs can offer high crystallinity, and so their structure can readily be characterized via diffraction. In addition, due to the presence of metal ions, and predefined molecular building blocks, which exhibit multiple chemical moieties, the potential energy surface for a guest molecule within a MOF cavity possesses multiple minima that correspond to well-defined adsorption sites with varying binding energies. This makes diffraction the most direct way to probe static site-specific binding properties. Given this, the focus of this work is on understanding the structure-derived function of MOFs of interest in gas separation/storage applications. To do this, standard adsorption experiments are coupled with in-situ diffraction techniques. The latter is able to reveal the location and orientation of adsorbed guest species inside the framework, provide insights into the framework response to varying pressure, temperature, and atmosphere, and allow one to determine the relative differences in binding energy between neighboring adsorption sites. The aforementioned information can then be used to rationalize the relationship between different physiochemical features of a given framework and their performance in an adsorption process. In addition to establishing structure-property relations for the usage of MOFs, the obtained experimental results are used to corroborate those calculated by DFT methods, providing a stress test for computational methods aimed at predicting the structures and properties of hypothetical MOFs. The first chapter of this thesis, the introduction, offers a brief review of different characterization methods for studying gas adsorption/separation applications in MOFs. Although the focus of this chapter is on carbon dioxide capture as the application, these characterization methods are also used to study other adsorption processes. The next three chapters, chapters 2-4, present several prominent MOF families that are of interest in hydrogen storage and carbon dioxide capture applications. These case studies demonstrate how using in-situ diffraction techniques coupled with adsorption measurements and DFT calculations can be used to gain molecular level insight into how small molecules bind inside the selected MOF structures. Further, these structure-property correlation studies are able to help unveil how altering MOF building blocks, such as metals and ligands, gives rise to enhanced or diminished adsorption properties. The last chapter, chapter 5, is a study of the response of an activated MOF to varying temperature; in this chapter various structural motions, which are responsible for a large negative thermal expansion (NTE), are elucidated.
Occurrence of cavitation in hydraulic machines is a challenging issue because it is often accompanied with loss of efficiency, noise emissions, vibrations, and erosion damages. Tip vortices, in particular, are an ideal site for the development of cavitation as the static pressure at their cores usually drops much below the freestream pressure. The first part of the present thesis is focused on the effect of dissolved gas content and other physical parameters on Tip Vortex Cavitation (TVC) trailing from an elliptical hydrofoil. The inception and desinence thresholds measured at different flow conditions reveal that TVC often disappears at cavitation indices higher than the inception thresholds. The measurements show that TVC desinence pressure increases with the gas content and may even reach to atmospheric pressure. This is explained by the convective diffusion of dissolved gases from water into the cavity due to local supersaturation. The extent of the delay in desinence is, however, dictated by the bulk flow parameters. Owing to flow visualizations, it is asserted that the formation of a laminar separation bubble at the hydrofoil tip is a necessary condition for the delayed desinence. The separation bubble acts like a shelter and creates a relatively calm area at the vortex core. The second part of the thesis is dedicated to TVC mitigation. First, the effectiveness of non-planar winglets in suppressing TVC is investigated. For this purpose, an elliptical hydrofoil is selected as the baseline and various winglets are realized by simply bending the last 5 or 10% of the span at ±45° and ±90° dihedral angles. TVC inception-desinence tests reveal undeniable advantages for the winglets while the hydrodynamic performances of the hydrofoils remain intact. It is observed that a longer winglet bent at a higher angle and toward the pressure side is more effective in TVC suppression. For instance, the 90°-bent-downward winglet reduces the TVC inception index from 2.5 for the baseline down to 0.8 at 15 m/s freestream velocity and 14° incidence angle. Stereo-PIV measurements show that for this winglet, the maximum tangential velocity of the tip vortex falls to almost half of the baseline and the axial velocity reduces tangibly at the vortex core. Second, the capacity of a flexible trailing thread in alleviating TVC is examined. To this end, one nylon thread with various diameters and lengths is attached to the tip of the elliptical hydrofoil. Under almost all the tested flow conditions, the thread experiences flutter due to hydro-elastic instabilities. Thereafter, the vortical flow forces the oscillating thread to coil around the vortex axis. This rotational motion decelerates the axial velocity at the vortex core due to increased drag and augmented turbulent mixing. A sufficiently thick thread may also be sucked into the vortex core under the effect of the pressure field. This results in the whipping motion, which consists of the quasi-periodic coincidence of a part of the thread and the tip vortex axis and is considerably superior to the rotational motion in TVC mitigation. It is shown analytically and confirmed experimentally that the whipping motion leads to viscous core thickening. Altogether, the results presented in this thesis provide a better understanding of the physics of TVC and pave the way for future designs with less vulnerability to cavitation.
Neoclassical tearing modes (NTMs), magnetic islands located at rational $q$ surfaces, are an important class of resistive magnetohydrodynamics (MHD) instabilities in tokamak plasmas, with $q$ the safety factor. NTMs are one of the main constraints of the achievable plasma pressure by increasing the local radial transport and NTMs can lead to plasma disruptions. It is therefore crucial to understand the physics of NTMs and ensure their reliable control. This thesis explores the physics and control of NTMs, by means of dedicated experiments in the TCV tokamak and interpretative simulations with the modified Rutherford equation (MRE), a model widely used in interpreting island width evolutions. Triggerless NTMs originating from unstable tearing modes (TMs, stability index $\Delta'>0$) and saturating under the effects of the perturbed bootstrap current are the main focus of this thesis. In TCV, triggerless NTMs are reproducibly observed in low-density discharges with strong near-axis electron cyclotron current drive (ECCD), providing an excellent opportunity of studying these modes. Instead of direct computations of $\Delta'$, a model for $\Delta'$ is developed based on extensive experiments and interpretative simulations. This model facilitates the clarification of the complete evolution of triggerless NTMs, from onset as TMs to saturation as NTMs. Our $\Delta'$ model also explains an unexpected density dependence of the onset of NTMs, where NTMs only occur with a certain range of density that broadens with increasing near-axis ECCD power and with lower plasma current. The density range is found to result from the density and plasma current dependence of the stability of ohmic plasmas and the density dependence of ECCD efficiency. Given its high localization and flexibility, off-axis ECH/ECCD will be used for NTM control in future tokamaks. Comprehensive experimental and numerical studies of the dynamics of NTMs are carried out in this thesis, concerning both the stabilization of existing NTMs and the prevention of NTMs by means of preemptive off-axis ECCD. It is shown and predicted that the prevention of NTMs is much more efficient than NTM stabilization in terms of EC power. Interpretative simulations of the complex set of experiments constrain well the coefficients in the MRE and quantify NTM evolutions. The prevention effects from off-axis ECCD are found to result from local ECH/ECCD instead of a change of $\Delta'$. A key element of a reliable real-time (RT) control of NTMs is the alignment of EC beams with the target mode location. A small sinusoidal sweeping of the deposition location of EC beams around the target location proves to be effective for both NTM stabilization and prevention, making it a promising technique. Integrated control of NTMs, plasma pressure and model-estimated $q$ profiles is demonstrated on TCV, including advanced plasma state reconstruction, monitoring, supervision and actuator management. A RT-capable MRE module, based on our validated MRE, is developed for the first time and tested by extensive offline simulations for TCV and AUG. It provides a more intelligent physics-based NTM controller, aware of the EC power it requires to stabilize or prevent a given NTM. The information from the RT-MRE is also valuable for RT actuator allocations and decision-making in view of the overall integrated control, in particular for future long-pulse tokamaks like ITER and DEMO.
An existence result is presented for the dynamical low rank (DLR) approximation for random semi-linear evolutionary equations. The DLR solution approximates the true solution at each time instant by a linear combination of products of deterministic and stochastic basis functions, both of which evolve over time. A key to our proof is to find a suitable equivalent formulation of the original problem. The so-called Dual Dynamically Orthogonal formulation turns out to be convenient. Based on this formulation, the DLR approximation is recast to an abstract Cauchy problem in a suitable linear space, for which existence and uniqueness of the solution in the maximal interval are established.
The global photovoltaic market is mostly dominated by solar cells based on crystalline silicon (c-Si), which are covering 95% of the market. This thesis concerns silicon heterojunction (SHJ), a high-efficiency technology with a 2% market share in 2018, but showing a steady global production capacity growth and potential for industry-compatible modules with a conversion efficiency above 21%. This thesis focuses on improving the device efficiency in real operating conditions. More specifically, different effects affecting a cell used in non-standard conditions are assessed, including different temperatures and illuminations, during prolonged exposure to light and the effect of surface inhomogeneities. The main SHJ technology specificity consists of the use of thin hydrogenated amorphous silicon (a-Si) layers that provide excellent c-Si surface passivation. This enables very high operating voltages compared to other c-Si-based technologies, which directly reduces the loss of efficiency induced by higher operating temperatures. The cell's properties are measured at very low temperatures are studied, which amplifies charges transport phenomena. The voltage shows a deviation from its theoretical curve, which we explain to be related to a decrease of the positive contact holes selectivity due to the influence of the transparent conductive oxide layer. An increase in the resistive losses at low temperature is measured. These losses are caused by a resistivity increase in the different a-Si layers and interfaces because of a thermal energy decrease at low temperatures. These effects may also affect cell performances at room temperature and under operating conditions. This implies that optimisation at standard test conditions (STC), generally used to assess the performance of PV modules, does not necessarily result in the same optimum as in real climatic conditions, which generally correspond to a higher temperature and/or a lower illumination. We show that by adding carbon in the front doped layer, a transparency increase, and thus a current gain is observed at the expense of larger ohmic losses, which overall reduces the efficiency at STC. Using simulations and real climate data, we show that using such a layer leads to a relative energy production gain of 0.5 to 0.8%, both in a moderate and arid climate. Another specificity of SHJ cells is a slight efficiency increase when exposed to light for a few days. We show that in case of a too-small positive doped layer charges reserve, cell performance degradation is observed instead of an improvement. In order to avoid such degradation, the positively doped layer should be sufficiently thick and doped to screen the influence of the subsequent transparent conductive oxide layer. A treatment in which a forward bias voltage is applied to the cell allows triggering the gain in efficiency without causing the detrimental effects of light exposure. Finally, the influence of a non-homogeneous surface is studied. Parasitic electric currents are induced between the well-passivated area and poor passivation area through the metallisation. Using dedicated samples, correspondence between these currents, and an often measured drop of the low injection lifetime is described and linked to the detrimental effect of either the sample edges or by surface defects. Concretely, these effects can affect the performances of the cell at low illumination, reducing the performances in real outdoor conditions.
Mood disorders, in particular depression, are a major burden of our society. Due to the poor knowledge of the biological basis of these diseases, classification remains based on arbitrary symptomatic parameters. As a result, the existing pharmacological treatments have difficulties targeting relevant pathophysiological processes leading to high level of non-responding patients. Magnetic resonance spectroscopy (MRS) provides an outstanding means of measuring biochemical processes in vivo and can help identifying metabolic pathways that are associated with a given pathological condition. In this thesis, we have taken advantage of state-of-the-art MRS technologies at high field for studying metabolic dysfunctions associated with behavioral impairments in animal models of mood disorder. The overall goal consisted in finding potential biomarkers and endophenotypes (i.e. heritable biomarkers) with MRS, associate them with a molecular/physiological mechanism and evaluate the effect of a treatment targeting the observed dysfunction. We have successfully identified neuroenergetic abnormalities in different limbic regions of the brain in two mouse models of mood disorders; with a genetic or an environmental origin. Genetic deletion of an important metabolic regulator in mouse brain led to hippocampal neuroenergetic impairment and susceptibility to environmental stressors. Treating the animals with ebselen, an energy boosting mood stabilizer, allowed us to reduce the animalâs sensitivity to stress. With the same approach, we observed energy-related biomarkers associated with susceptibility to stress in the nucleus accumbens of genetically identical mice. We found that social hierarchy can predict the response to a chronic stressor and that behavioral impairments could be prevented by administering an energy stimulating compound, acetyl-L-carnitine. Finally, in an additional project, we have used MRS in an embryonic model in ovo to investigate for markers related to metabolic remodeling during neurogenesis. Our results support the idea that mood disorders arise from energy metabolism fragility in different regions of the limbic system with both environmental and genetic origin. Due to the high translational potential of MRS into clinics, our findings provide new biological targets or routes to study for a better understanding of mood disorders.
Urbanization intensifies as a global trend, exposing transportation networks to ever increasing levels of congestion. As network usage increases with available infrastructure, building new roads is not a solution. Design of intelligent transportation systems, involving identification, estimation, and feedback control methods with dynamical traffic models, is emerging as a feasible way to improve operation of existing infrastructure. Nevertheless, complexity of large-scale networks, spatiotemporal propagation of congestion, and uncertainty in traveler choices present considerable challenges for modeling, estimation, and control of road transport systems. This dissertation focuses on development of novel and practicable optimization-based traffic control and estimation methods for improving mobility in large-scale urban road networks. Part 1 is dedicated to identification, estimation, and control methods based on macroscopic traffic dynamics for perimeter controlled urban networks. Obtaining accurate estimates of model parameters and traffic states is critical for feedback perimeter control systems. In chapter 2, a nonlinear moving horizon estimation (MHE) scheme is proposed for combined state and demand estimation for a two-region urban network with dynamical modeling via macroscopic fundamental diagram (MFD). A traffic control framework consisting of identification, state estimation, and control methods is developed in chapter 3, enabling model-based feedback perimeter control of city-scale traffic. Part 2 focuses on traffic management methods considering regional route guidance. Equipping traffic controllers with route guidance carries potential for high performance congestion management. Chapter 4 contains model predictive control (MPC) schemes integrating route guidance and perimeter control actuators, capable of superior performance compared to using only perimeter control. A hierarchical traffic controller is designed in chapter 5, employing a path assignment mechanism to realize macroscopic route guidance commands of a network-level MPC.
All mechanical systems, naturally occurring or human-produced, are subjected to friction and wear at the interface of solid constituents. Large portions of energy dissipation and loss of material, in every-day life and industrial applications alike, are due to friction and wear. Mitigating their effects could save between 1\% and 2\% of the GDP of a developed country. Some systems governed by friction and wear can have an even more important bearing on human lives, such as earthquakes nucleating from the sliding of tectonic faults. Despite the tremendous impact of tribological phenomena on society, their understanding has remained empirical, and to this day no predictive model has emerged. Interface processes such as friction and wear are difficult to investigate because of the large number of underlying physical phenomena (e.g. adhesion, fracture, etc.) and the difficulty of observing them at contact interfaces. Although research endeavors into friction and wear have not produced predictive models, they have identified key components of tribological systems necessary to build such models. Central among them is the idea that solids may not be in contact across their apparent interface area, but instead a much smaller "true contact area." This true contact area is the result of the surfaces in contact being inevitably rough. In addition, contact pressures on roughness peaks are expected to cause plastic flow of material, drastically changing the properties of the contact interface, and the role it plays in tribological processes. Therefore, the aim of this PhD thesis is to develop tools for the modeling of elastic-plastic rough contact interfaces, and to study the applicability of knowledge of the contact state to the modeling of interface phenomena. The first part of this objective is the development of a novel computational approach to volume integral methods, which are used to solve elastic-plastic rough surface contact. Volume integral methods have the advantage over the finite-element method in that they can represent exactly elastic constitutive behavior and semi-infinite bodies, which are commonly used in rough contact applications. This thesis develops a new fundamental solution used in a volume integral approach, which drastically improves computation times and required memory over previous approaches. Derived directly in the Fourier domain, this fundamental solution makes optimal use of the fast-Fourier transform while retaining the advantages of classical volume integral methods. In the second part, this numerical approach is used to study the so called "Archard's wear coefficient", and to up-scale known micro-scale adhesive wear mechanisms to the macro-scale via rough contact simulations. These show that wear is an emergent process dependent on the interaction of micro-scale mechanisms: they demonstrate the role of plastic deformations in the crack nucleation process, and the necessity to look beyond the true contact area to understand tribological phenomena. While this thesis remains quite fundamental, the tools and codes developed can be used outside the realm of elastic-plastic contact, and the up-scaling approach to wear that we have established is a first step towards predictive models.